Climatology of rainfall observed from satellite and surface data in the Mediterranean basin

Remote Sensing and Geographic Information Systems for Design and Operation of Water Resources Systems (Proceedings of Rabat Symposium S3, April 1997). IAHS Publ. no. 242, 1997 165 Climatology of rainfall observed from satellite and surface data in the Mediterranean basin LUCA G. LANZA & FRANCO SICCARDI Institute of Hydraulics, University ofgenova, 1 Montallegro, 1-16129 Genova, Italy Abstract Some research issues emerging from the work undertaken in the framework of a recent EU-funded project, named ACROSS, are presented and discussed. The project aims at producing a unified climatology of rainfall over the eastern Mediterranean region remotely sensed over sea, using passive microwave imagery, and ground observed over land. From the analysis and interpretation of polar orbiting satellite data of the sea surface it appears that within the coastal region, the estimates of rainfall amounts are affected by large uncertainties due to the differences in the microwave spectral emission of the sea and land surface. For a narrow strip along the coast, a.boundary region where algorithms produce large uncertainties in the results, a linking problem is observed. The two basic datasets were implemented within a relational database and a hydrologically oriented geographic information system (GIS). INTRODUCTION: THE ACROSS PROJECT A recent project developed under the Avicenne Programme of the European Community, named ACROSS (Analysed Climatology of Rainfall Observed from Satellite and Surface data) is the framework where the research described in this paper is being presently developed and discussed. The project aims at producing a unified climatology of rainfall over the eastern Mediterranean region remotely sensed over sea and ground observed over land. Such a database will allow a much more satisfactory evaluation and interpretation than it has been possible hitherto of overland and over-water anomalies. The study covers the period from 1978 to the present day since microwave satellite observation is available. The study area includes a window region with boundaries at 12 W, 29 N and 45 E, 48 N covering the entire Mediterranean basin from the Atlantic in the west to the Middle East countries and extending north to include the Black Sea region. The partners of this project are the Hydraulic Institute of the University of Genoa (Italy), which is the Co-ordinator, the Centre for Remote Sensing of the University of Bristol (UK), the Civil Engineering Department of the Middle East Technical University (Turkey), the Department of Geology of the University of Amman (Jordan) and ICARDA (International Centre for Agricultural Research in the Dry Areas) is participating with its Syrian agency. This paper is presented by the Project Co-ordinator and includes contributions from all the involved institutions as presented in their interim reports and/or during workshops held in recent project meetings. An overview of the project tasks and objectives is given synthetically in Table 1. The work undertaken during the first phase of the project was mainly devoted to the

166 Luca G. Lanza & Franco Siccardi Table 1 Overview of the ACROSS project: tasks and objectives. Work Area 1: Collecting ground-based rainfall data and development of remote sensing techniques and graphical tools Task 1.1 Collection ofraingauge data from WMO stations and local Months 1 to 6 networks in the study area at a fine resolution. Task 1.2 Development and calibration of techniques for processing SMMR Months 1 to 6 and SSM/I images for rainfall and surface characteristics. Task 1.3 Acquisition and preparation of satellite images over the established Months 1 to 6 window over the Mediterranean area for the period 1978-1994. Task 1.4 Development of graphical tools for data analysis and presentation. Months 1 to 6 Work Area 2: Application of techniques for SMMR data analysis and development of techniques for integration with land and sea surface data Task 2.1 Processing of SMMR images over the established window within the Months 6 to 12 Task 2.2 Mediterranean area for the period 1978-1987. Development of a GIS-based system capable of storing, geo- Months 6 to 12 referencing and integrating satellite derived and ground-based data. Work Area 3: Application of techniques for SSM/I data analysis and development of techniques for integration with land and sea surface data Task 3.1 Processing of SSM/I images over the established window within the Months 12 to 18 Mediterranean area for the period 1978-1987. Task 3.2 Integration of satellite and ground-based data for land and sea Months 12 to 18 surfaces. Work Area 4: Presentation and description of results Task 4.1 Preparation and production of climatological atlases of rainfall and Months 18 to 24 surface characteristics, along with description notes. Task 4.2 Generation of time series analyses of rainfall and associated Months 18 to 24 surface characteristics, and presentation of results. acquisition of the basic data resources needed for the eventual development of dedicated research studies. In particular the collection, collation and analysis of raingauge data for suitable and available climatic stations in the study area were addressed, in order to produce maps of average rainfall for months, seasons and/or years, plus maps, graphs and statistics for rainfall variability and departures from the norm. A large number of raingauge and meteo-climatic stations were selected in the study area and the collection of data for the period 1978-1994 had already started. In particular daily rainfall data series from the NOAA NCDC dataset were acquired and complemented with sparse data from the national networks as provided by each partner institution for its own country and for neighbouring ones (Boni & Lanza, 1996; Altinbilek, 1996; Salameh, 1996; Oweis & Goebel, 1996) in order to achieve the information in a density of about one raingauge per 625 km 2 (25 X 25 km grid). At the same time the collection and geo-registration of SMMR (Scanning Multichannel Microwave Radiometer) (1978-1987) and SSM/I (Special Sensor Micro wave/imager) (1987-1994) images was addressed together with their preliminary analysis for rainfall over water within the study area to complement the above mentioned products, and to complete the regional picture for the eastern Mediterranean region (Todd et al., 1996a). The definition of suitable graphical tools for the presentation of project results was also addressed and the acquisition of ancillary data was completed. The two datasets were implemented within a relational database and a hydrologically oriented

Climatology of rainfall observed from satellite and surface data in the Mediterranean basin 167 SATELLITE MAPS PRE PROCESSING ft RAINGAUGE TIME SERIES PRE PROCESSING SATELLITE DERIVED RAINFALL MAPS RAINGAUGE DERTVED RAINFALL MAPS ism* ANCILLARYDATA DATA INTEGRATION THE ACROSS CLIMATOLOGY _ OUTPUT DIGITAL =*_-=?^V PRODUCTS OUTPUT MAPS - monthly totals B-_ - monthly mean ^f Fig. 1 The GIS structure for the ACROSS project (Lanza, 1996). geographic information system (GIS). The GIS structure for the ACROSS project is depicted in Fig. 1 (Lanza, 1996). The 30" Digital Elevation Model of the Mediterranean region, obtained from USGS, was selected as the basic information over which both satellite and raingauge data are displayed (see Fig. 2(b)). THE TWO DIFFERENT DATASETS As for the raingauge information, the consortium agreed that daily time series should be collected from a suitable number of meteorological and climatic stations within the study area, for the countries involved, and in particular from those stations located along the coast or in a very close distance to the sea. Stations along the coast

168 Luca G. Lanza & Franco Siccardi Fig. 2 (a) Sample over-water precipitation map for the study area (Todd et ai, 1996a,b) and (b) grey scale representation of the 30" DEM obtained from USGS. especially those located at small islands are indeed very important because they allow an interpolation between the two datasets (the traditional and the satellite derived) and even the calibration and verification of rainfall estimates over the sea surface. Specific attention is thus devoted to the coastal areas where the two datasets must fit to each other to provide a final unified rainfall climatology of the region. Hourly data from selected raingauge stations are collected for a series of case studies, e.g. in the case of precipitation peaks, when they are detected by the analysis of daily data. Therefore raingauge data are stored in the ACROSS archives as hourly/daily time series associated with a given location which is characterized by two geographical co-ordinates, in a linear latitude/longitude reference system and an elevation value. Data from the SMMR (Scanning Multichannel Microwave Radiometer) were obtained (Todd et al., 1996b) in the form of Temperature Calibrated Tapes (TCTs) for the operation period from 25 October 1978 to 20 August 1987 from the US National Space Science Data Center (NSSDS), Greenbelt, Maryland. The raw radiometric readings were corrected for actual antenna patterns including sidelobe effects, and the vertical and horizontal polarisation components of the brightness temperature were separated. SSM/I data were obtained in three formats: (a) the Wentz Format directly obtained from Frank Wentz of Remote Sensing Systems, Santa Rosa, California; (b) the NESDIS IB Format obtained from the Marshall Space Flight Center (MSFC) at Huntsville, USA; and

Climatology of rainfall observedfrom satellite and surface data in the Mediterranean basin 169 (c) the FNMOC Format obtained from the MSFC too. Due to the huge amount of data involved in ACROSS, including both SSM/I and SMMR, an automatic quality control has been incorporated into the data processing procedure (Todd et al., 1996b). The first result was the complete acquisition and preparation of satellite imagery in order to enable the production of rainfall estimates from remotely sensed data. This was made (Todd et al., 1996a,b) for the whole study period (1978-1994) for both the Scanning Multichannel Microwave Radiometer (SMMR) and the Special Sensor Microwave/Imager (SSM/I). Data processing was also performed so as to apply quality control routines which discard bad data, using the available automated codes in the case of SMMR images and incorporating a manual quality control scheme in the case of SSM/I data where corrupt scan lines leading to errors in rainfall estimates are removed without affecting the non-corrupted data. The calibration and validation of selected algorithms for estimation of rainfall over water using passive microwave satellite data was carried out (Todd et al, 1996a,b), using radar rainfall maps. Finally, passive microwave algorithms for the retrieval of land surface information (soil moisture, vegetation index) were analysed in order to propose a suitable strategy for monitoring surface characteristics (Todd et al, 1996a). A sample for an over-water precipitation map is shown in Fig. 2(a) (Todd et al., 1996a,b). THE LINKING PROBLEM From the analysis and interpretation of polar orbiting satellite data from sea surface it appears that within the coastal region, the estimates of rainfall amounts are affected by large uncertainties. Differences in the microwave spectral emission of the sea and land surface affect the beam filling near the shore: in fact different rainfall retrieval algorithms perform well either on sea or on land, but not on both sides. A narrow strip along the coast is therefore the boundary where algorithms produce large uncertainties in the results the linking problem. Rainfall amounts in this region have to be derived from joint observations from satellite systems and raingauge stations taking into account the very specific characteristics of the area and the physics of the processes in supporting the linking mathematical procedures: variations in rainfall regimes and in the distribution of anomalies are expected when moving from the sea towards the land surface. These variations are enhanced in some regions by the presence of steep and high altitude mountain ranges near to the coastline. Attention is thus being paid to the analysis of methods for producing suitable confidence intervals to be associated with both raingauge and satellite rainfall measurements. It is a challenging task to produce a unified climatological map from the two different datasets, where raingauge rainfall is obtained by accumulated daily totals while satellite rainfall measurements are evaluated by interpolating sample images that are available only twice a day. Finally the two datasets are characterized by different uncertainty sources and the confidence limits associated with them are fairly different from each other. For the satellite dataset, two kinds of errors are expected: the bias due to instrument errors (which is usually corrected during the calibration step, thanks to

170 Luca G. Lanza & Franco Siccardi the empirical relationship that is operationally used for climatological rainfall retrieval) and the sampling error due to the scarce availability of satellite passages (which also depends on the latitude of the study area). In the ACROSS project bias errors will be neglected and the attention is focused on sampling errors. Huffman et al. (1996) justify this assumption on the basis of the observation that sampling errors are usually dominant in monthly precipitation estimates at climatological scales. The functional representation of sampling errors proposed by Huffman et al. (1996) in terms of the error variance of the average over a finite set of observations is inversely proportional to the number of independent samples and directly proportional to a parameterization of the conditional precipitation rate. This gives a chance to quantify the uncertainty associated with satellite rainfall estimates also in terms of a "quality index" whose units are defined as "equivalent gauges", i.e. the approximate number of gauges required to produce the same error for a given rain total. As for the raingauge dataset, optimal interpolation methods such as kriging are based on the error variance minimization criterion and give analytical expressions for the minimum variance achieved after interpolation. The analytical expression allows the transformation of single site uncertainty estimates into some confidence limits for the isohyets. So the two datasets can be compared in terms of confidence limits, and the linking problem in the coastal zone is addressed through a minimum variance criterion again. Because of the huge amount of data required and the need of providing a unified climatological map for the whole Mediterranean region, definitive results are not yet available for the ACROSS project whose duration is actually two years and which started on July 1995. However the linking problem between the two main datasets of rainfall measurements available for climatological studies is a stimulating research issue. It holds several implications in both practical applications and further research studies. This paper is presented in the spirit of encouraging the discussion and cooperation in research among hydrologists, climatologists and remote sensing scientists from several countries bordering the Mediterranean Sea, within and outside the ACROSS project, towards a better understanding of over-land and over-water precipitation anomalies and rainfall distribution. CONCLUSIONS The ACROSS project is providing a good opportunity for international co-operation in research in the Mediterranean region allowing the integration of rainfall patterns over land and water in a single map on a time-determinate basis. This new product will be useful in many contexts, including climatological and agricultural applications, desertification studies, and advanced research in climatology and meteorology. In this case it will provide means of exploring the meteorological causes of relevant anomalies over land, which are triggered by extreme events originated over extended sea surfaces. This is of paramount importance in the Mediterranean region where extreme events produce showers whose duration do not exceed a few hours but whose rainfall depths exceeds somewhere between 1/5 and 1/3 of the total annual average precipitation.

Climatology of rainfall observed from satellite and surface data in the Mediterranean basin 171 Acknowledgement The work presented in this paper was supported by the Commission of the European Communities, DGXI, Avicenne Initiative, in the framework of the contract AVI080, "ACROSS: Analysed Climatology of Rainfall Observed from Satellite and Surface data in the Mediterranean Area". REFERENCES Altinbilek, D. (1996) First Year Report for the ACROSS Project (Middle East Technical Univ.). The European Boni, G. & Lanza, L. (1996) Collection of raingauge data from WMO stations and local networks. First Semester Report for the ACROSS Project (Univ. of Genoa). The European Huffman, G. J., Adler, R. F., Arkin, P., Chang, A., Ferraro, R., Gruber, A., Janowiak, J., McNab, A., Rudolf, B. & Schneider, U. (1996) The Global Precipitation Climatology Project (GPCP), combined precipitation data set. Submitted to Bull. Am. Met. Soc.. Lanza, L. (1996) Development of a GIS-based system capable of storing, georeferencing and integrating satellite-derived information and ground-based data. First Year Report for the ACROSS Project (Univ. of Genoa). The European Oweis, T. & Goebel, W. (1996) First Year Report for the ACROSS Project (ICARDA). The European Commission, DGXI, AVICENNE Initiative. Salameh, E. (1996) The climate of Jordan. First Year Report for the ACROSS Project (Univ. of Jordan). The European Todd, M. C, Barrett, E. C. & Beaumont, M. J. (1996a) Development and calibration of techniques for processing SMMR and SMM/I data for estimation of rainfall and land surface characteristics. First Year Report for the ACROSS Project (Univ. of Bristol). The European Todd, M. C, Barrett, E. C. & Beaumont, M. J. (1996b) Processing of SMMR images for rainfall over the Mediterranean Sea from 1978-1987. First Year Report for the ACROSS Project (Univ. of Bristol). The European